Random Forest Feature Importance

In the random forest feature importance method, a random forest model is trained on the data points. A random forest model trains machine learning models referred to as decision trees. The decision trees create splits in the dataset based on patterns to form a branching hierarchy in which data points can be assigned. Once each tree is trained, the results are averaged together to arrive at a prediction. The following diagram shows how decision trees work in the random forest model.

Using the random forest feature importance method, the features contributing the most to the decision splits are identified. The features that contribute the most to the best and worst predicted cases are ranked and ordered in terms of importance, and the features are chosen accordingly.